Hi Marsha,
> Could someone tell me if there are any objective ways to decide
> when to
> 'unwarp and realign' fMRI data rather than just 'realign'?
Not really, no.
Realign will remove all (well most) "first-order" effects of movements, i.e. those that can be explained by a rigid-body model where all movement occurrs bewteen frames.
Unwarp will, in addition to first order effects, attempt to correct for the apparent shape/size changes that occurr when the B0 field changes as a result of the changes in subject position.
In general you can say that if movements are small ( < 1degree rotation) the rigid-body model is "good enough", and Unwarp removes very little. For slightly larger movements Unwarp will start to remove a sizeable amount of movement-related variance. Of course this will also depend on a the degree of distortion to begin with. On a crappy old scanner with large distortions the effects may be apparent already a ~0.5 degrees rotation, whereas on a new scanner with fast gradients (and possibly combined with GRAPPA or some other suitable acronym) movements may be considerably larger before this effect is noticed.
I do think it is a good idea to be consistent within a study. I.e. don't use Unwarp on some and not on others.
Good luck Jesper
|